Hossein Raghfar; Narges Ajorlo
Abstract
The purpose of this study is to calculate Value at Risk (VaR) of a selection of bank's currency portfolio, using GARCH-EVT-Copula (GEC) approach. Today's main challenge of a banking system is to calculate and quantify the risks that the system is encountered. There are numerous approaches ...
Read More
The purpose of this study is to calculate Value at Risk (VaR) of a selection of bank's currency portfolio, using GARCH-EVT-Copula (GEC) approach. Today's main challenge of a banking system is to calculate and quantify the risks that the system is encountered. There are numerous approaches to calculate the risks. Usually these approaches assume a common known distribution for the assets portfolio and generally a normal distribution is utilized for the experimental models. Nevertheless, the distributions of the assets are fat-tailed distribution and consequently normal distribution assumption may lead to inaccurate estimation. This article does not assume a specific asset distribution. This study applies autoregressive threshold variances (GJR-GARCH) for intertemporal individual's asset variable returns distribution. It also utilizes extreme value theory or the fat-tailed distributions and Coppola functions for all asset returns in an asset portfolio. In this study VaR is estimated using variance-covariance and historical simulation methods. Finally, in order to test the reliability of the applied models Kopic method is used. The sample data of the bank's currency portfolio consists of the market daily figures of the US Dollar, Japan's Yen, Turkish Lire, Emirate Dirham, Korean Won, and Euro exchange rates from March 21, 2007 till April 19, 2012. The results showed that the estimated VaR using GEC model is higher than the one estimated using the other two methods. They also show that reliability and precision of Kopic test is higher than those of variance-covariance and historical simulation models.
mirhossein mousavi; Hossein Raghfar; Mansooreh Mohseni
Volume 18, Issue 54 , April 2013, , Pages 119-152
Abstract
The traditional approaches for estimating VAR assume that the joint distribution is well-known and the most commonly used normality of the joint distribution of the assets return. In reality, the financial asset return distribution has fatter tails than normal distributions. On the other hand, the use ...
Read More
The traditional approaches for estimating VAR assume that the joint distribution is well-known and the most commonly used normality of the joint distribution of the assets return. In reality, the financial asset return distribution has fatter tails than normal distributions. On the other hand, the use of linear correlation to model the dependence structure shows many disadvantages. Therefore, the problem raised from normality could lead to an inadequate VaR estimate. In order to overcome these problems, this paper resorts to the copula theory which allows the joint distribution of the portfolio to be free from any normality and linear correlation. Combining copula and the forecast function of the GARCH model, this paper proposes a new method, called conditional copula-GARCH, to compute the VaR of portfolios. Examined data in this study includes daily price of selected portfolio, composed of 17 equities for 1082 days in Tehran stock exchange. Presented model compared with traditional methods (including the historical simulation method & variance_covariance method). the results show that conditional copula-GARCH model captures the VaR more successfully at 95% confidence.